Use the timestamps directly as the ringbuffer index. We can save some
conversions to bytes and there is a direct mapping to RTP timestamp,
clock position and ringbuffer index.
Simplify the source a little. Remove the buffering state, we always
start with read and write pointers separted by the target buffering.
Always write to index of the clock, not just for the first packet.
This ensure the timestamp on the packet always matches the time it was
processed in the graph.
Always use the timestamp of the graph clock for RTP packets. Add an
option to apply a random or fixed offset.
Add a ts-refclk option on the sender to specify a reference clock to
use for timestamping. This will activate the direct timestamp mode and
signal this in the SDP.
Parse ts-refclk and ts-offset from the SDP. Make it possible to match
them in rules. Add option to activate the direct-timestamp, where the
rate matching is disabled and the timestamps are used directly for
writing and reading from the ringbuffer using the graph clock.
This makes it possible to set a PTP clock on sink and source and avoid
rate matching by using a shared clock.
Add support for using other clocks.
clock.id can be used to set one of the system clocks.
clock.device can be used to open a clock device such as a PTP clock
device.
Use a dll to track the progress of non-monotonic clocks.
Use the control changes to load a new IR, then signal to data thread to
use the new IR. In the data thread, signal the main thread to free the old
IR after we swapped it out.
THis moves all allocations from the data thread into the main thread.
SOFA is a file format used for storing and accessing spatial audio data, namely head-related transfer functions. These can be used to create binaural spatial sound using head- or earphones.
This commit introduces libmysofa as an optional dependency for loading SOFA files and creates a spatializer plugin for the filter-chain
ci: install libmysofa-devel for full build
ci: bump FDO_DISTRIBUTION_TAG
For lower than default rates, limit the upsampling to 2 times the default
rate. This avoid 8000Hz to the upsampled to 192000 when a 44100 is
available (but not 48000).
For higher rates limit the upsampling to *3.
Try to upsample to something in the same rate family so that the amount
of resample filter is minimal.
If that doesn't work, try to downsample to something in the same rate
family above the 44100 threshold.
If that also doesn't work, downsample to the highest available samplerate.
When the driver has no followers to run, we can keep the current rate
and don't need to switch. This avoids a useless switch to the default
samplerate when all followers are removed.
In theory, the pipewiresrc element is a live element by default.
However, that is not reflected in code, as the element never sets
that flag unless explicltly requested through stream properties.
Make it live by default, but still respect if consumers of the
element mark it as not live.
Try to keep half the packet size in the ringbuffer as well. This helps
us adapt to the packet size of the sender.
Drop samples from the ringbuffer for the first packet we read. This
makes us lock onto the stream with the exact requested latency.
